{"id":"https://openalex.org/W4412945258","doi":"https://doi.org/10.18653/v1/2025.acl-long.768","title":"Masks Can be Learned as an Alternative to Experts","display_name":"Masks Can be Learned as an Alternative to Experts","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412945258","doi":"https://doi.org/10.18653/v1/2025.acl-long.768"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.768","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.768","pdf_url":"https://aclanthology.org/2025.acl-long.768.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.acl-long.768.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5119204489","display_name":"Peiyu Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Peiyu Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119204490","display_name":"Tianwen Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tianwen Wei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119204491","display_name":"Bo Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bo Zhu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119204492","display_name":"Xin Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xin Zhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5119204493","display_name":"Shuicheng Yan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shuicheng Yan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.27481117,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"15800","last_page":"15811"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10672","display_name":"Design Education and Practice","score":0.06960000097751617,"subfield":{"id":"https://openalex.org/subfields/2210","display_name":"Mechanical Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10672","display_name":"Design Education and Practice","score":0.06960000097751617,"subfield":{"id":"https://openalex.org/subfields/2210","display_name":"Mechanical Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12124","display_name":"Engineering Education and Curriculum Development","score":0.05990000069141388,"subfield":{"id":"https://openalex.org/subfields/2214","display_name":"Media Technology"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10809","display_name":"Occupational Health and Safety Research","score":0.05739999935030937,"subfield":{"id":"https://openalex.org/subfields/3614","display_name":"Radiological and Ultrasound Technology"},"field":{"id":"https://openalex.org/fields/36","display_name":"Health Professions"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.607650876045227}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.607650876045227}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.768","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.768","pdf_url":"https://aclanthology.org/2025.acl-long.768.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.768","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.768","pdf_url":"https://aclanthology.org/2025.acl-long.768.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412945258.pdf","grobid_xml":"https://content.openalex.org/works/W4412945258.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"In":[0],"this":[1,49,124],"work,":[2],"we":[3,51,100],"investigate":[4],"how":[5],"to":[6,26,37,103],"sparsify":[7,61],"a":[8,15,133],"pre-trained":[9],"dense":[10,119],"large":[11],"language":[12],"model":[13,54,150],"into":[14],"mixture-of-experts":[16],"(MoE)":[17],"architecture":[18],"for":[19,29,78,91,136,147],"faster":[20],"inference.Our":[21],"approach":[22,65,82],"applies":[23],"mask":[24],"matrix":[25],"the":[27,39,53,114,145],"activations":[28],"each":[30],"expert,":[31],"constrained":[32],"by":[33],"L":[34],"0":[35],"regularization":[36],"minimize":[38],"number":[40],"of":[41,95,113,126],"activated":[42,117],"parameters.To":[43],"ensure":[44],"minimal":[45],"performance":[46,79,105],"loss":[47],"under":[48],"constraint,":[50],"initialize":[52],"with":[55,110],"all":[56],"parameters":[57,116],"active":[58],"and":[59,87,140],"progressively":[60],"it":[62],"during":[63],"training.This":[64],"proves":[66],"more":[67,92,138],"efficient":[68,93],"than":[69],"oneshot":[70],"sparsification":[71],"techniques,":[72],"which":[73],"typically":[74],"require":[75],"significant":[76],"resources":[77],"recovery.Moreover,":[80],"our":[81],"automatically":[83],"identifies":[84],"shared,":[85],"token-specific,":[86],"inactive":[88],"experts,":[89],"allowing":[90],"allocation":[94],"computational":[96,128],"resources.Through":[97],"extensive":[98],"experiments,":[99],"achieve":[101],"up":[102],"97%":[104],"retention":[106],"on":[107],"downstream":[108],"tasks":[109],"only":[111],"50%":[112],"feed-forward":[115],"in":[118],"models.Beyond":[120],"improving":[121],"inference":[122],"efficiency,":[123],"strategy":[125],"sharing":[127],"units":[129],"among":[130],"experts":[131],"provides":[132],"principled":[134],"foundation":[135],"building":[137],"scalable":[139],"generalizable":[141],"MoE":[142],"architectures,":[143],"paving":[144],"way":[146],"future":[148],"expert-based":[149],"designs.":[151]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}